Nvidia admits $100 billion OpenAI deal remains unsigned, stoking AI bubble concerns

Nvidia OpenAI Deal Remains Uncommitted Two Months After Announcement

The Nvidia OpenAI deal, initially announced in September as a landmark partnership worth up to $100 billion, has not progressed beyond a letter of intent, Nvidia CFO Colette Kress confirmed at the UBS Global Technology and AI Conference in Arizona 1

1

2

2

. "We still haven't completed a definitive agreement, but we're working with them," Kress stated, addressing investor questions about the framework of what CEO Jensen Huang once described as "the biggest AI infrastructure project in history" 3

3

. The revelation is notable given that the announcement helped fuel an AI infrastructure rally in September, sending Nvidia shares up 4% at the time 3

3

.

The original plan outlined deploying millions of Nvidia GPUs over several years, backed by at least 10 gigawatts of data center capacity—enough to power more than 8 million U.S. homes 4

4

. Nvidia pledged to invest up to $100 billion in OpenAI as each tranche comes online, with analysts estimating the deal could generate as much as $500 billion in revenue for the chipmaker 3

3

. Despite the lack of a signed agreement, Kress emphasized that Nvidia's relationship with OpenAI remains "a very strong partnership" spanning more than a decade, with OpenAI considering Nvidia its "preferred partner" for compute 3

3

.

AI Infrastructure Challenges and Forecasting Demand Complexities

Nvidia's latest 10-Q filing reveals why the $100 billion OpenAI deal has not yet finalized, highlighting significant risk factors associated with long-term partnerships 1

1

. The company states plainly that "there is no assurance that we will enter into definitive agreements with respect to the OpenAI opportunity or other potential investments, or that any investment will be completed on expected terms, if at all" 1

1

. The filing emphasizes that AI data center infrastructure deployment depends heavily on the availability of data center capacity, energy, and capital—all of which face regulatory, technical, and construction challenges 3

3

.

Forecasting demand has become increasingly complex as Nvidia moves to an annual cadence of architecture releases, including Hopper, Blackwell, and Vera Rubin systems 3

3

. The company must order GPUs, HBM memory, networking gear, and other components more than a year in advance, often via non-cancelable, prepaid contracts 3

3

. If customers scale back, delay financing, or change direction, Nvidia warns it may end up with excess inventory, cancellation penalties, or inventory provisions 3

3

. Kress noted that the roughly $500 billion of Blackwell and Vera Rubin system demand for 2025-2026 does not include any OpenAI demand relating to the potential deal 1

1

.

Investor Concerns About AI Bubble and Circular Deals

The admission that the Nvidia OpenAI deal is not yet finalized adds to intensifying discussion around investor concerns about an AI bubble and circular deals in the AI ecosystem 4

4

5

5

. Over the past year, Nvidia has struck a series of deals with AI startups and invested in firms that are also major customers, creating a landscape where companies become dependent on each other 1

1

. Nvidia recently announced plans to commit up to $10 billion to OpenAI rival Anthropic, with Kress stating this deal could also add to Nvidia's $500 billion in chip bookings 4

4

.

Large institutions like the Bank of England have warned about "sharp market corrections" when this bubble bursts, outlining potential risks for tech companies tied into long-term partnerships 1

1

. Despite these concerns, Nvidia shares rose around 2.6% following Colette Kress's remarks at the conference 1

1

4

4

. However, the stock remains down more than 11% over the last month amid concern about competition 2

2

.

Nvidia Defends Market Position Against Google TPUs Competition

Addressing competition concerns, Colette Kress was emphatic that Nvidia is "absolutely not" worried about companies like Google making their own TPUs 1

1

2

2

. "Everybody is on our platform," she noted, suggesting it would take a long time for the industry to shift away from Nvidia hardware 1

1

. Nvidia shares slipped last week on reports that Meta was in talks with Google about using Google's custom chips, called TPUs, in its data centers 2

2

. In response, Nvidia took to social media to defend its GPUs, saying on X that its popular chips were "a generation ahead of the industry" 2

2

.

Kress reiterated that Nvidia envisions $3 trillion to $4 trillion in global AI investments by the end of the decade, emphasizing that the economy is "in the early parts" of transitioning to the AI data center infrastructure needed to power AI applications 2

2

. OpenAI currently purchases through cloud partners like Microsoft and Oracle rather than through the new direct arrangement laid out in the letter of intent, though Kress noted that OpenAI "does want to go direct" once a definitive agreement is reached 3

3

.